Phase 7d-b — `@vectorized` / `@no_vectorize` array SIMD gates¶

Summary¶

@no_vectorize on a def forces scalar loops for ArrayDotF64 / ArrayBinOpF64; @vectorized(lanes=4) is validated and documents intent; lanes=8 is a compile error (E0322). @vectorized may prefix for loops (parse only).

Agent continuation¶

Read MirFn.no_vectorize, EmitCtx.enable_array_simd, check_proc_decorators.
Run ./li-tests/run_all.sh (expect 170 pass).
Then elaborate loop-body @vectorized to MIR (not just parse); @parallel + math loops.
Blocked on f64x8 codegen for lanes=8.

Changed¶

Path	What
`compiler/mir/`	`MirDecorator.lanes`, `MirFn.no_vectorize`
`compiler/codegen/emit.cpp`	`enable_array_simd` gate
`compiler/types/policy_module.cpp`	E0322 lanes policy
`compiler/parser/parser.cpp`	`@… for` parse
`li-tests/decorators/vectorized_*.li`	four tests

Not changed¶

Loop-body lowering from @vectorized (metadata on for only).
@parallel on math @ / matmul.
f64x8 vectors.

Breaking¶

N/A.

Security¶